15-05-2012, 11:53 AM
Speech For The Disabled
IMECS2009_pp567-572.pdf (Size: 708.14 KB / Downloads: 110)
INTRODUCTION
Various kinds of speech recognition systems are available today but there are hardly a few which help the ‘deaf-and dumb’. In the recent years, due to rapid industrialization there has been a rapid increase in the number of speech-disabled victims due to oral diseases, accidents, etc...
Whatever the cause may be, they are rendered helpless and unable to communicate with the outside world! Hence it’s high time that we not only focus on our development, but also theirs! We require speech recognition software. For this, the required input for the speech recognition software can be either a written input in the form of text or voice input (which uses similarities in sounds).
Cepstral Analysis-
Here let x(n) be a voiced speech signal, e(n) is the excitation function and h(n) is the vocal tract impulse response . The excitation function and the vocal tract impulse response are convoluted to produce speech signal,
x(n)=e(n)*h(n);
De-convolution is carried out by a general non-linear filtering method referred to as homographic filtering. In this method, the convolution operation is converted into addition giving the output known as complex cepstrum.
Linear prediction analysis.
•Powerful speech processing technique
•Here, sample values of speech can be approximated as a linear combination of past p samples
•An output of a program written using concepts of DSP is shown. It uses the algorithm of linear prediction analysis,
The object of linear prediction is to form a model of a linear time-invariant digital system through observation of input and output sequences. That is, to estimate a set of coefficients which will describe the behaviour of an LTI system when its design is not available to us and we cannot choose what input to present.